Dear all, please, join me discussing my latest preprint.
Feel free to leave a review as well.
Conversational Artificial Intelligence (CAI) was the latest big step towards artificial general intelligence (AGI): models of AI that are generic, that can solve problems without being restricted to a small set of problems they were designed for. OpenAI with chatGPT gave a big step towards this direction. On the other hand, being able to classify snakes from images can help the general public, and even experts to classify snakes (e.g., finding the right snake serum). There are more than 2.000 different species of snakes worldwide. On this paper, we explore the latest releases of openAI APIs. We integrate with a previous work we did, using transfer learning to classify snakes. We have used their GPT Vision API for describing uploaded images textually, their latest chatGPT API for deciding which model to use for classifying an image and also to generate a human-friendly final textual response. Additionally, we also tested their GPT 4 API instead of the chatGPT API as a comparison. We found that the integration was a success. By giving our models to openAI API, it decides which model to call as so it could give an educated guess to the user based on an uploaded image from the snake. The best result was when we used the GPT 4 API, something we already expected. We hope that this prototype can be scaled up with more species. We did a test of concept, and showed that the basics needed to build such a system for snake classification is possible with current technologies. We hope to contribute to sparking research that brings together computer scientists and biologists, generally called bioinformatics.
=
Join me on this discussion:
SnakeChat: a conversational-AI based app for snake classification
https://www.qeios.com/read/Y13B20
- Badges
- Science topic
- Similar topics
- Correction